fix: robust PaddleOCR init with multiple fallback strategies

PaddleOCR 3.x removed show_log param and lang='latin'. Try multiple
init strategies in order until one succeeds.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

.gitea/workflows/ci.yaml

@@ -140,20 +140,117 @@ jobs:
           python -m pytest tests/bqas/ -v --tb=short || true
 
   # ========================================
-  # Deploy via Coolify (nur main, kein PR)
+  # Build & Deploy auf Hetzner (nur main, kein PR)
   # ========================================
 
-  deploy-coolify:
-    name: Deploy
+  deploy-hetzner:
     runs-on: docker
     if: github.event_name == 'push' && github.ref == 'refs/heads/main'
     needs:
       - test-go-consent
-    container:
-      image: alpine:latest
+    container: docker:27-cli
     steps:
-      - name: Trigger Coolify deploy
+      - name: Deploy
         run: |
-          apk add --no-cache curl
-          curl -sf "${{ secrets.COOLIFY_WEBHOOK }}" \
-            -H "Authorization: Bearer ${{ secrets.COOLIFY_TOKEN }}"
+          set -euo pipefail
+          DEPLOY_DIR="/opt/breakpilot-core"
+          COMPOSE_FILES="-f docker-compose.yml -f docker-compose.hetzner.yml"
+          COMMIT_SHA="${GITHUB_SHA:-unknown}"
+          SHORT_SHA="${COMMIT_SHA:0:8}"
+          REPO_URL="${GITHUB_SERVER_URL}/${GITHUB_REPOSITORY}.git"
+
+          # Services die deployed werden
+          SERVICES="postgres valkey qdrant minio ollama mailpit embedding-service rag-service backend-core consent-service health-aggregator"
+
+          echo "=== BreakPilot Core Deploy ==="
+          echo "Commit: ${SHORT_SHA}"
+          echo "Deploy Dir: ${DEPLOY_DIR}"
+          echo "Services: ${SERVICES}"
+          echo ""
+
+          # 1. Repo auf dem Host erstellen/aktualisieren via Helper-Container
+          echo "=== Updating code on host ==="
+          docker run --rm \
+            -v "${DEPLOY_DIR}:${DEPLOY_DIR}" \
+            --entrypoint sh \
+            alpine/git:latest \
+            -c "
+              if [ ! -d '${DEPLOY_DIR}/.git' ]; then
+                echo 'Erstmaliges Klonen nach ${DEPLOY_DIR}...'
+                git clone '${REPO_URL}' '${DEPLOY_DIR}'
+              else
+                cd '${DEPLOY_DIR}'
+                git fetch origin main
+                git reset --hard origin/main
+              fi
+            "
+          echo "Code aktualisiert auf ${SHORT_SHA}"
+
+          # 2. .env sicherstellen
+          docker run --rm -v "${DEPLOY_DIR}:${DEPLOY_DIR}" alpine \
+            sh -c "
+              if [ ! -f '${DEPLOY_DIR}/.env' ]; then
+                echo 'WARNUNG: ${DEPLOY_DIR}/.env fehlt!'
+                echo 'Erstelle .env aus .env.example mit Defaults...'
+                if [ -f '${DEPLOY_DIR}/.env.example' ]; then
+                  cp '${DEPLOY_DIR}/.env.example' '${DEPLOY_DIR}/.env'
+                  echo '.env aus .env.example erstellt'
+                else
+                  echo 'Kein .env.example gefunden — Services starten mit Defaults'
+                fi
+              else
+                echo '.env vorhanden'
+              fi
+            "
+
+          # 3. Shared Network erstellen (falls noch nicht vorhanden)
+          docker network create breakpilot-network 2>/dev/null || true
+
+          # 4. Build + Deploy via Helper-Container
+          echo ""
+          echo "=== Building + Deploying ==="
+          docker run --rm \
+            -v /var/run/docker.sock:/var/run/docker.sock \
+            -v "${DEPLOY_DIR}:${DEPLOY_DIR}" \
+            -w "${DEPLOY_DIR}" \
+            docker:27-cli \
+            sh -c "
+              set -e
+              COMPOSE_FILES='-f docker-compose.yml -f docker-compose.hetzner.yml'
+
+              echo '=== Building Docker Images ==='
+              docker compose \${COMPOSE_FILES} build --parallel \
+                backend-core consent-service rag-service embedding-service health-aggregator
+
+              echo ''
+              echo '=== Starting infrastructure ==='
+              docker compose \${COMPOSE_FILES} up -d postgres valkey qdrant minio mailpit
+
+              echo 'Warte auf DB + Cache...'
+              sleep 10
+
+              echo ''
+              echo '=== Starting Ollama + pulling bge-m3 ==='
+              docker compose \${COMPOSE_FILES} up -d ollama
+              sleep 5
+
+              # bge-m3 Modell pullen (nur beim ersten Mal ~670MB)
+              echo 'Pulling bge-m3 model (falls noch nicht vorhanden)...'
+              docker exec bp-core-ollama ollama pull bge-m3 2>&1 || echo 'WARNUNG: bge-m3 pull fehlgeschlagen (wird spaeter nachgeholt)'
+
+              echo ''
+              echo '=== Starting application services ==='
+              docker compose \${COMPOSE_FILES} up -d \
+                embedding-service rag-service backend-core consent-service health-aggregator
+
+              echo ''
+              echo '=== Health Checks ==='
+              sleep 15
+              for svc in bp-core-postgres bp-core-valkey bp-core-qdrant bp-core-ollama bp-core-embedding-service bp-core-rag-service bp-core-backend bp-core-consent-service bp-core-health; do
+                STATUS=\$(docker inspect --format='{{.State.Status}}' \"\${svc}\" 2>/dev/null || echo 'not found')
+                echo \"\${svc}: \${STATUS}\"
+              done
+            "
+
+          echo ""
+          echo "=== Deploy abgeschlossen: ${SHORT_SHA} ==="

.gitea/workflows/deploy-coolify.yml

@@ -7,12 +7,10 @@ on:
 
 jobs:
   deploy:
-    runs-on: docker
-    container: alpine:latest
+    runs-on: ubuntu-latest
     steps:
       - name: Deploy via Coolify API
         run: |
-          apk add --no-cache curl
           echo "Deploying breakpilot-core to Coolify..."
           HTTP_STATUS=$(curl -s -o /dev/null -w "%{http_code}" \
             -X POST \

docker-compose.coolify.yml

@@ -143,7 +143,7 @@ services:
       - breakpilot-network
 
   # =========================================================
-  # OCR SERVICE (PaddleOCR PP-OCRv5)
+  # OCR SERVICE (PaddleOCR PP-OCRv5 Latin)
   # =========================================================
   paddleocr-service:
     build:
@@ -157,12 +157,10 @@ services:
       FLAGS_use_mkldnn: "0"
     volumes:
       - paddleocr_models:/root/.paddleocr
-    labels:
-      - "traefik.http.services.paddleocr.loadbalancer.server.port=8095"
     deploy:
       resources:
         limits:
-          memory: 6G
+          memory: 4G
     healthcheck:
       test: ["CMD", "curl", "-f", "http://127.0.0.1:8095/health"]
       interval: 30s

paddleocr-service/main.py

@@ -1,4 +1,4 @@
-"""PaddleOCR Remote Service — PP-OCRv4 on x86_64 (CPU)."""
+"""PaddleOCR Remote Service — PP-OCRv5 Latin auf x86_64."""
 
 import io
 import logging
@@ -27,22 +27,31 @@ def _load_model():
         logger.info("Importing paddleocr...")
         from paddleocr import PaddleOCR
 
-        logger.info("Loading PaddleOCR model (PP-OCRv4, lang=en)...")
-        _engine = PaddleOCR(
-            lang="en",
-            use_angle_cls=True,
-            show_log=False,
-            enable_mkldnn=False,
-            use_gpu=False,
-        )
-        logger.info("PaddleOCR model loaded — running warmup...")
-        # Warmup with tiny image to trigger any lazy init
-        dummy = np.ones((30, 100, 3), dtype=np.uint8) * 255
-        _engine.ocr(dummy)
+        logger.info("Import done. Loading PaddleOCR model...")
+        # Try multiple init strategies for different PaddleOCR versions
+        inits = [
+            # PaddleOCR 3.x (no show_log)
+            dict(lang="en", ocr_version="PP-OCRv5", use_angle_cls=True),
+            # PaddleOCR 3.x with show_log
+            dict(lang="en", ocr_version="PP-OCRv5", use_angle_cls=True, show_log=False),
+            # PaddleOCR 2.8+ (latin)
+            dict(lang="latin", use_angle_cls=True, show_log=False),
+            # PaddleOCR 2.8+ (en, no version)
+            dict(lang="en", use_angle_cls=True, show_log=False),
+        ]
+        for i, kwargs in enumerate(inits):
+            try:
+                _engine = PaddleOCR(**kwargs)
+                logger.info(f"PaddleOCR init succeeded with strategy {i}: {kwargs}")
+                break
+            except Exception as e:
+                logger.info(f"PaddleOCR init strategy {i} failed: {e}")
+        else:
+            raise RuntimeError("All PaddleOCR init strategies failed")
         _ready = True
-        logger.info("PaddleOCR ready to serve")
+        logger.info("PaddleOCR model loaded successfully — ready to serve")
     except Exception as e:
-        logger.error(f"Failed to load PaddleOCR: {e}", exc_info=True)
+        logger.error(f"Failed to load PaddleOCR model: {e}")
 
 
 @app.on_event("startup")
@@ -50,14 +59,15 @@ def startup_load_model():
     """Start model loading in background so health check passes immediately."""
     global _loading
     _loading = True
-    threading.Thread(target=_load_model, daemon=True).start()
+    thread = threading.Thread(target=_load_model, daemon=True)
+    thread.start()
     logger.info("Model loading started in background thread")
 
 
 @app.get("/health")
 def health():
     if _ready:
-        return {"status": "ok", "model": "PP-OCRv4"}
+        return {"status": "ok", "model": "PP-OCRv5-latin"}
     if _loading:
         return {"status": "loading"}
     return {"status": "error"}
@@ -78,30 +88,25 @@ async def ocr(
     img = Image.open(io.BytesIO(img_bytes)).convert("RGB")
     img_np = np.array(img)
 
-    try:
-        result = _engine.ocr(img_np)
-    except Exception as e:
-        logger.error(f"OCR failed: {e}", exc_info=True)
-        raise HTTPException(status_code=500, detail=f"OCR failed: {e}")
-
-    if not result or not result[0]:
-        return {"words": [], "image_width": img_np.shape[1], "image_height": img_np.shape[0]}
+    result = _engine.ocr(img_np)
 
     words = []
-    for line in result[0]:
+    for line in result[0] or []:
         box, (text, conf) = line[0], line[1]
         x_min = min(p[0] for p in box)
         y_min = min(p[1] for p in box)
         x_max = max(p[0] for p in box)
         y_max = max(p[1] for p in box)
-        words.append({
-            "text": str(text).strip(),
-            "left": int(x_min),
-            "top": int(y_min),
-            "width": int(x_max - x_min),
-            "height": int(y_max - y_min),
-            "conf": round(float(conf) * 100, 1),
-        })
+        words.append(
+            {
+                "text": text.strip(),
+                "left": int(x_min),
+                "top": int(y_min),
+                "width": int(x_max - x_min),
+                "height": int(y_max - y_min),
+                "conf": round(conf * 100, 1),
+            }
+        )
 
     return {
         "words": words,

paddleocr-service/requirements.txt

@@ -1,5 +1,5 @@
-paddlepaddle>=2.6.0,<3.0.0
-paddleocr>=2.7.0,<3.0.0
+paddlepaddle>=3.0.0
+paddleocr>=2.9.0
 fastapi>=0.110.0
 uvicorn>=0.25.0
 python-multipart>=0.0.6